Google Launches Gemini 3.1 Ultra With 2-Million Token Context and Built-In Code Execution

Google has released Gemini 3.1 Ultra, a major leap in multimodal AI capability that features a 2-million token context window and a new sandboxed Code Execution tool that lets the model write, run, and test code mid-conversation.

Multimodal by Default

Gemini 3.1 Ultra works natively across text, image, audio, and video inputs simultaneously. This means users can upload a video, ask questions about its content, have the model analyze audio quality, and generate code to process the footage — all in a single conversation thread.

The 2-million token context window doubles the previous generation’s capacity and sets a new industry benchmark. For perspective, this is enough to process approximately 20 full-length novels or an entire year’s worth of corporate email communications.

Code Execution Changes the Game

The integrated Code Execution feature allows Gemini 3.1 Ultra to not just generate code, but execute it in a sandboxed environment during the conversation. This enables real-time data analysis, chart generation, mathematical verification, and iterative debugging without leaving the chat interface.

Google has made Gemini 3.1 Ultra available through Google AI Studio, Vertex AI, and as part of the Google One AI Premium subscription at $19.99/month. Enterprise pricing for Vertex AI deployments starts at custom rates based on usage volume.